FIELD: physics.
SUBSTANCE: in the method, a user terminal gallery is obtained; identification and marking of the image are carried out; a training sample set is received; each of a plurality of training image sequences is inserted; the characteristic coefficients are taught between the levels of hidden nodes of the model for determining the original image of the survey plan; a test sample set is received; identification of test images is carried out; the classification accuracy of the model for determining the image survey plan is determined; if the classification accuracy is less than the given threshold value, the following is performed: updating the training sample set; training, in accordance with the updated training sample set, the characteristic coefficients between the corresponding levels of the hidden nodes of the model for determining the image survey plan; performing an iteration of updating the model for determining the image survey plan; performing, in accordance with the updated training sample set, a test for the classification accuracy of the updated model for determining the image survey plan.
EFFECT: improving the accuracy of image classification.
11 cl, 14 dwg
Authors
Dates
2017-09-29—Published
2015-12-28—Filed